Cilj identifikacije procesa je na temelju skupa mjernih podataka (ulaznih i izlaznih signala), pronaći strukturu i parametre odgovarajućeg matematičkog modela koji najbolje opisuje promatrani proces. Struktura se najčešće određuje kroz teorijsku analizu ili nasumičnim odabirom, dok se parametri određuju nekom od metoda za procjenu parametara. Važno je dobro procijeniti dinamiku procesa kako bi se odabrao odgovarajući red modela i/ili mrtvo vrijeme. Ako se struktura procesa odabere velikog reda (s puno parametara) onda proces uči krive podatke, odnosno počinje modelirati i šum. Pri identifikaciji parametara modela važno je odrediti vektor parametara tako da odstupanje u vladanju matematičkog modela od vladanja procesa bude što manje, u stacionarnom i dinamičkom režimu rada. U radu su obrađene najčešće korištene strukture modela procesa: AR, ARX, FIR, ARMAX, OE i BJ model. Isto tako obrađene su metode procjene parametara: LS metoda, IV metoda i ML metoda (direktna i rekurzivna rješenja). Za odabrani proces (sustav kuglica-greda) provedena je parametarska identifikacija koristeći programski paket MATLAB i njegov alat System Identification Toolbox.

The purpose of system identification is find a strucuture and parameters of mathematical model which best describes observed process, based on a set of measured data (inputs and outputs). The structure is usually determined through theoretical analysis or random selection, while parameters are determining with some of methods for parameter estimation. It is important to good estimate the dynamics of the process in order to select the appropriate rank of the model and/or dead time. If the structure of the process is selected with a large order (many parameters), then the process is learning wrong information, begins to model noise also. Also very important is to determine vector of parameters so that deviation between conduct of mathematical model and conduct of real system be as small as possible in stationary and dynamic mode. In this paper are processed the most commonly used model structures: AR, ARX, FIR, ARMAX, OE and BJ model. Also parameter estimation methods that have been processed are: LS method, IV method and ML method (direct and recursive solutions). Parametric system identification was conducted for selected process (ball and beam system) using software package MATLAB and its tool System Identification Toolbox.